Training Wasserstein GANs without gradient penalties
This addresses the challenge of stable training for generative adversarial networks, offering a computationally efficient alternative without hyperparameter tuning, though it appears incremental as it builds on existing WGAN frameworks.
The paper tackles the problem of training Wasserstein GANs without gradient penalties by proposing a stable method using c-transform objectives from optimal transport theory, which effectively enforces Lipschitz constraints and yields competitive synthetic image generators on MNIST, F-MNIST, and CIFAR-10 datasets.
We propose a stable method to train Wasserstein generative adversarial networks. In order to enhance stability, we consider two objective functions using the $c$-transform based on Kantorovich duality which arises in the theory of optimal transport. We experimentally show that this algorithm can effectively enforce the Lipschitz constraint on the discriminator while other standard methods fail to do so. As a consequence, our method yields an accurate estimation for the optimal discriminator and also for the Wasserstein distance between the true distribution and the generated one. Our method requires no gradient penalties nor corresponding hyperparameter tuning and is computationally more efficient than other methods. At the same time, it yields competitive generators of synthetic images based on the MNIST, F-MNIST, and CIFAR-10 datasets.